Face Generation¶

In this project, you'll use generative adversarial networks to generate new images of faces.

Get the Data¶

You'll be using two datasets in this project:

MNIST
CelebA

Since the celebA dataset is complex and you're doing GANs in a project for the first time, we want you to test your neural network on MNIST before CelebA. Running the GANs on MNIST will allow you to see how well your model trains sooner.

If you're using FloydHub, set data_dir to "/input" and use the FloydHub data ID "R5KrjnANiKVhLWAkpXhNBe".

data_dir = './data'

# FloydHub - Use with data ID "R5KrjnANiKVhLWAkpXhNBe"
#data_dir = '/input'


"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper

helper.download_extract('mnist', data_dir)
helper.download_extract('celeba', data_dir)

Downloading mnist: 9.92MB [00:03, 3.17MB/s]                                                                            
Extracting mnist: 100%|████████████████████████████████████████████████████████| 60.0K/60.0K [00:39<00:00, 1.52KFile/s]
Downloading celeba: 1.44GB [04:31, 5.31MB/s]

Extracting celeba...

Explore the Data¶

MNIST¶

As you're aware, the MNIST dataset contains images of handwritten digits. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
import os
from glob import glob
from matplotlib import pyplot

mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'mnist/*.jpg'))[:show_n_images], 28, 28, 'L')
pyplot.imshow(helper.images_square_grid(mnist_images, 'L'), cmap='gray')

<matplotlib.image.AxesImage at 0x17aa8195048>

CelebA¶

The CelebFaces Attributes Dataset (CelebA) dataset contains over 200,000 celebrity images with annotations. Since you're going to be generating faces, you won't need the annotations. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'img_align_celeba/*.jpg'))[:show_n_images], 28, 28, 'RGB')
pyplot.imshow(helper.images_square_grid(mnist_images, 'RGB'))

<matplotlib.image.AxesImage at 0x17aa84cad30>

Preprocess the Data¶

Since the project's main focus is on building the GANs, we'll preprocess the data for you. The values of the MNIST and CelebA dataset will be in the range of -0.5 to 0.5 of 28x28 dimensional images. The CelebA images will be cropped to remove parts of the image that don't include a face, then resized down to 28x28.

The MNIST images are black and white images with a single color channel while the CelebA images have 3 color channels (RGB color channel).

Build the Neural Network¶

You'll build the components necessary to build a GANs by implementing the following functions below:

model_inputs
discriminator
generator
model_loss
model_opt
train

Check the Version of TensorFlow and Access to GPU¶

This will check to make sure you have the correct version of TensorFlow and access to a GPU

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf

# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer.  You are using {}'.format(tf.__version__)
print('TensorFlow Version: {}'.format(tf.__version__))

# Check for a GPU
if not tf.test.gpu_device_name():
    warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
    print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))

TensorFlow Version: 1.3.0

C:\Users\Anna\Anaconda3\envs\py36\lib\site-packages\ipykernel_launcher.py:14: UserWarning: No GPU found. Please use a GPU to train your neural network.

Input¶

Implement the model_inputs function to create TF Placeholders for the Neural Network. It should create the following placeholders:

Real input images placeholder with rank 4 using image_width, image_height, and image_channels.
Z input placeholder with rank 2 using z_dim.
Learning rate placeholder with rank 0.

Return the placeholders in the following the tuple (tensor of real input images, tensor of z data)

import problem_unittests as tests

def model_inputs(image_width, image_height, image_channels, z_dim):
    """
    Create the model inputs
    :param image_width: The input image width
    :param image_height: The input image height
    :param image_channels: The number of image channels
    :param z_dim: The dimension of Z
    :return: Tuple of (tensor of real input images, tensor of z data, learning rate)
    """
    # TODO: Implement Function
    r_input  = tf.placeholder(tf.float32, (None,image_width, image_height, image_channels), name='r_input')
    z_input = tf.placeholder(tf.float32, (None, z_dim), name='z_input')
    learning_r = tf.placeholder(dtype=tf.float32, name='learning_r')
    return r_input , z_input, learning_r


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_inputs(model_inputs)

ERROR:tensorflow:==================================
Object was never used (type <class 'tensorflow.python.framework.ops.Operation'>):
<tf.Operation 'assert_rank_2/Assert/Assert' type=Assert>
If you want to mark it as used call its "mark_used()" method.
It was originally created here:
['File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\runpy.py", line 193, in _run_module_as_main\n    "__main__", mod_spec)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\runpy.py", line 85, in _run_code\n    exec(code, run_globals)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\ipykernel_launcher.py", line 16, in <module>\n    app.launch_new_instance()', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\traitlets\\config\\application.py", line 658, in launch_instance\n    app.start()', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\ipykernel\\kernelapp.py", line 477, in start\n    ioloop.IOLoop.instance().start()', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\zmq\\eventloop\\ioloop.py", line 177, in start\n    super(ZMQIOLoop, self).start()', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\tornado\\ioloop.py", line 888, in start\n    handler_func(fd_obj, events)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\tornado\\stack_context.py", line 277, in null_wrapper\n    return fn(*args, **kwargs)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\zmq\\eventloop\\zmqstream.py", line 440, in _handle_events\n    self._handle_recv()', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\zmq\\eventloop\\zmqstream.py", line 472, in _handle_recv\n    self._run_callback(callback, msg)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\zmq\\eventloop\\zmqstream.py", line 414, in _run_callback\n    callback(*args, **kwargs)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\tornado\\stack_context.py", line 277, in null_wrapper\n    return fn(*args, **kwargs)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\ipykernel\\kernelbase.py", line 283, in dispatcher\n    return self.dispatch_shell(stream, msg)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\ipykernel\\kernelbase.py", line 235, in dispatch_shell\n    handler(stream, idents, msg)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\ipykernel\\kernelbase.py", line 399, in execute_request\n    user_expressions, allow_stdin)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\ipykernel\\ipkernel.py", line 196, in do_execute\n    res = shell.run_cell(code, store_history=store_history, silent=silent)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\ipykernel\\zmqshell.py", line 533, in run_cell\n    return super(ZMQInteractiveShell, self).run_cell(*args, **kwargs)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\IPython\\core\\interactiveshell.py", line 2698, in run_cell\n    interactivity=interactivity, compiler=compiler, result=result)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\IPython\\core\\interactiveshell.py", line 2808, in run_ast_nodes\n    if self.run_code(code, result):', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\IPython\\core\\interactiveshell.py", line 2862, in run_code\n    exec(code_obj, self.user_global_ns, self.user_ns)', 'File "<ipython-input-44-69494a02a5a7>", line 22, in <module>\n    tests.test_model_inputs(model_inputs)', 'File "C:\\Users\\Anna\\Documents\\DeepLearning\\projects_delivered\\P5-face-generation\\problem_unittests.py", line 12, in func_wrapper\n    result = func(*args)', 'File "C:\\Users\\Anna\\Documents\\DeepLearning\\projects_delivered\\P5-face-generation\\problem_unittests.py", line 68, in test_model_inputs\n    _check_input(learn_rate, [], \'Learning Rate\')', 'File "C:\\Users\\Anna\\Documents\\DeepLearning\\projects_delivered\\P5-face-generation\\problem_unittests.py", line 34, in _check_input\n    _assert_tensor_shape(tensor, shape, \'Real Input\')', 'File "C:\\Users\\Anna\\Documents\\DeepLearning\\projects_delivered\\P5-face-generation\\problem_unittests.py", line 20, in _assert_tensor_shape\n    assert tf.assert_rank(tensor, len(shape), message=\'{} has wrong rank\'.format(display_name))', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\tensorflow\\python\\ops\\check_ops.py", line 617, in assert_rank\n    dynamic_condition, data, summarize)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\tensorflow\\python\\ops\\check_ops.py", line 571, in _assert_rank_condition\n    return control_flow_ops.Assert(condition, data, summarize=summarize)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\tensorflow\\python\\util\\tf_should_use.py", line 175, in wrapped\n    return _add_should_use_warning(fn(*args, **kwargs))', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\tensorflow\\python\\util\\tf_should_use.py", line 144, in _add_should_use_warning\n    wrapped = TFShouldUseWarningWrapper(x)', 'File "C:\\Users\\Anna\\Anaconda3\\envs\\py36\\lib\\site-packages\\tensorflow\\python\\util\\tf_should_use.py", line 101, in __init__\n    stack = [s.strip() for s in traceback.format_stack()]']
==================================
Tests Passed

Discriminator¶

Implement discriminator to create a discriminator neural network that discriminates on images. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "discriminator" to allow the variables to be reused. The function should return a tuple of (tensor output of the discriminator, tensor logits of the discriminator).

def discriminator(images, reuse=False, alpha=0.2):
    """
    Create the discriminator network
    :param images: Tensor of input image(s)
    :param reuse: Boolean if the weights should be reused
    :return: Tuple of (tensor output of the discriminator, tensor logits of the discriminator)
    """
    # TODO: Implement Function
    with tf.variable_scope('discriminator', reuse=reuse):
        l1 = tf.layers.conv2d(images, 64, 5, strides=2, padding='same')
        relu1 = tf.maximum(alpha * l1, l1)
        # 16x16x64
        
        l2 = tf.layers.conv2d(relu1, 128, 5, strides=2, padding='same')
        l2 = tf.layers.batch_normalization(l2, training=True)
        relu2 = tf.maximum(alpha * l2, l2)
        # 8x8x128
        
        l3 = tf.layers.conv2d(relu2, 256, 5, strides=2, padding='same')
        l3 = tf.layers.batch_normalization(l3, training=True)
        relu3 = tf.maximum(alpha * l3, l3)
        # 4x4x256
        
        flat = tf.reshape(relu3, (-1, 4*4*256))
        #dropout_layer = tf.layers.dropout(inputs=flat, rate=dropout)
        logits = tf.layers.dense(flat, 1)
        out = tf.sigmoid(logits)

        
    return out, logits

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_discriminator(discriminator, tf)

Tests Passed

Generator¶

Implement generator to generate an image using z. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "generator" to allow the variables to be reused. The function should return the generated 28 x 28 x out_channel_dim images.

def generator(z, out_channel_dim, is_train=True, alpha=0.2):
    """
    Create the generator network
    :param z: Input z
    :param out_channel_dim: The number of channels in the output image
    :param is_train: Boolean if generator is being used for training
    :return: The tensor output of the generator
    """
 

    with tf.variable_scope('generator', reuse= not is_train):
        l1 = tf.layers.dense(z, 3*3*512)
        l1 = tf.reshape(l1, (-1, 3, 3, 512))
        l1 = tf.layers.batch_normalization(l1, training=is_train)
        l1 = tf.maximum(alpha * l1, l1)
        # 3x3x512
        
        l2 = tf.layers.conv2d_transpose(l1, 256, kernel_size=4, strides=2, padding='same')
        l2 = tf.layers.batch_normalization(l2, training=is_train)
        l2 = tf.maximum(alpha * l2, l2)
        # 6x6x256
        
        l3 = tf.layers.conv2d_transpose(l2, 128, kernel_size=4, strides=2, padding='valid')
        l3 = tf.layers.batch_normalization(l3, training=is_train)
        l3 = tf.maximum(alpha * l3, l3)
        #14x14x128
        
        # Output layer
        logits = tf.layers.conv2d_transpose(l3, out_channel_dim, kernel_size=5,strides=2, padding='same')
        out = tf.tanh(logits)
        # 28x28x5
        
        return out
    

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_generator(generator, tf)

Tests Passed

Loss¶

Implement model_loss to build the GANs for training and calculate the loss. The function should return a tuple of (discriminator loss, generator loss). Use the following functions you implemented:

discriminator(images, reuse=False)
generator(z, out_channel_dim, is_train=True)

def model_loss(input_real, input_z, out_channel_dim, alpha=0.2):
    """
    Get the loss for the discriminator and generator
    :param input_real: Images from the real dataset
    :param input_z: Z input
    :param out_channel_dim: The number of channels in the output image
    :return: A tuple of (discriminator loss, generator loss)
    """
    g_model = generator(input_z, out_channel_dim, alpha=alpha)
#    print(input_real.shape)
    d_model_real, d_logits_real = discriminator(input_real, reuse=False,alpha=alpha)
    d_model_fake, d_logits_fake = discriminator(g_model, reuse=True, alpha=alpha)

    d_loss_real = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_real, labels=tf.ones_like(d_model_real)))
    d_loss_fake = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, labels=tf.zeros_like(d_model_fake)))
    g_loss = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, labels=tf.ones_like(d_model_fake)))

    d_loss = d_loss_real + d_loss_fake

    return d_loss, g_loss    


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_loss(model_loss)

Tests Passed

Optimization¶

Implement model_opt to create the optimization operations for the GANs. Use tf.trainable_variables to get all the trainable variables. Filter the variables with names that are in the discriminator and generator scope names. The function should return a tuple of (discriminator training operation, generator training operation).

def model_opt(d_loss, g_loss, learning_rate, beta1):
    """
    Get optimization operations
    :param d_loss: Discriminator loss Tensor
    :param g_loss: Generator loss Tensor
    :param learning_rate: Learning Rate Placeholder
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :return: A tuple of (discriminator training operation, generator training operation)
    """
    # TODO: Implement Function
    
    # Get weights and bias to update
    t_vars = tf.trainable_variables()
    d_vars = [var for var in t_vars if var.name.startswith('discriminator')]
    g_vars = [var for var in t_vars if var.name.startswith('generator')]

    # Optimize
    with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
        d_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(d_loss, var_list=d_vars)
        g_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(g_loss, var_list=g_vars)

    return d_train_opt, g_train_opt
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_opt(model_opt, tf)

Tests Passed

Neural Network Training¶

Show Output¶

Use this function to show the current output of the generator during training. It will help you determine how well the GANs is training.

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np

def show_generator_output(sess, n_images, input_z, out_channel_dim, image_mode):
    """
    Show example output for the generator
    :param sess: TensorFlow session
    :param n_images: Number of Images to display
    :param input_z: Input Z Tensor
    :param out_channel_dim: The number of channels in the output image
    :param image_mode: The mode to use for images ("RGB" or "L")
    """
    cmap = None if image_mode == 'RGB' else 'gray'
    z_dim = input_z.get_shape().as_list()[-1]
    example_z = np.random.uniform(-1, 1, size=[n_images, z_dim])

    samples = sess.run(
        generator(input_z, out_channel_dim, False),
        feed_dict={input_z: example_z})

    images_grid = helper.images_square_grid(samples, image_mode)
    pyplot.imshow(images_grid, cmap=cmap)
    pyplot.show()

Train¶

Implement train to build and train the GANs. Use the following functions you implemented:

model_inputs(image_width, image_height, image_channels, z_dim)
model_loss(input_real, input_z, out_channel_dim)
model_opt(d_loss, g_loss, learning_rate, beta1)

Use the show_generator_output to show generator output while you train. Running show_generator_output for every batch will drastically increase training time and increase the size of the notebook. It's recommended to print the generator output every 100 batches.

def train(epoch_count, batch_size, z_dim, learning_rate, beta1, get_batches, data_shape, data_image_mode):
    """
    Train the GAN
    :param epoch_count: Number of epochs
    :param batch_size: Batch Size
    :param z_dim: Z dimension
    :param learning_rate: Learning Rate
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :param get_batches: Function to get batches
    :param data_shape: Shape of the data
    :param data_image_mode: The image mode to use for images ("RGB" or "L")
    """
    # TODO: Build Model
    #saver = tf.train.Saver()
    #sample_z = np.random.uniform(-1, 1, size=(72, z_size))

    steps = 0
   
    print_step = 10
    show_step = 100
    alpha = 0.2
    images_to_show = 9
    losses =[]

    _, width, height, channels = data_shape
    print(data_shape)
    
    r_input, z_input, learning_r = model_inputs(width, height, channels, z_dim)
    print('model input evaluated')
    print(r_input.shape)
    d_loss, g_loss = model_loss(r_input, z_input, channels, alpha = alpha)
    d_opt, g_opt = model_opt(d_loss, g_loss, learning_r, beta1)
    
    print('training...')
    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
        for epoch_i in range(epoch_count):
            for batch_images in get_batches(batch_size):
                steps += 1
                batch_images = 2*batch_images
                # Random Noise as input
                z_batch = np.random.uniform(-1, 1, size=(batch_size, z_dim))
                _ = sess.run(d_opt, feed_dict={r_input : batch_images, z_input: z_batch, learning_r: learning_rate})
                # Let the generator run twice, so that the discriminator does not converge that fast.
                _ = sess.run(g_opt, feed_dict={z_input: z_batch, r_input : batch_images, learning_r: learning_rate})
                
                if steps % print_step == 0:
                    train_loss_d = d_loss.eval({z_input: z_batch, r_input: batch_images})
                    train_loss_g = g_loss.eval({z_input: z_batch})
                    
                    print("Step: {}:".format(steps),
                          "Epoch {}/{}...".format(epoch_i+1, epoch_count),
                          "Discriminator Loss: {:.4f}...".format(train_loss_d),
                          "Generator Loss: {:.4f}".format(train_loss_g))
                    losses.append((train_loss_d, train_loss_g))
                
                if steps % show_step == 0:
                    show_generator_output(sess, images_to_show, z_input, channels, data_image_mode)
        #saver.save(sess, './checkpoints/generator.ckpt')
    print("Training finished")

MNIST¶

Test your GANs architecture on MNIST. After 2 epochs, the GANs should be able to generate images that look like handwritten digits. Make sure the loss of the generator is lower than the loss of the discriminator or close to 0.

batch_size = 32
z_dim = 200
learning_rate = 0.0002
beta1 = 0.3


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 2

mnist_dataset = helper.Dataset('mnist', glob(os.path.join(data_dir, 'mnist/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, mnist_dataset.get_batches,
          mnist_dataset.shape, mnist_dataset.image_mode)

(60000, 28, 28, 1)
model input evaluated
(?, 28, 28, 1)
training...
Step: 10: Epoch 1/2... Discriminator Loss: 4.5246... Generator Loss: 0.0141
Step: 20: Epoch 1/2... Discriminator Loss: 1.5945... Generator Loss: 3.2732
Step: 30: Epoch 1/2... Discriminator Loss: 0.8160... Generator Loss: 2.1300
Step: 40: Epoch 1/2... Discriminator Loss: 1.9401... Generator Loss: 0.1872
Step: 50: Epoch 1/2... Discriminator Loss: 0.6848... Generator Loss: 1.5699
Step: 60: Epoch 1/2... Discriminator Loss: 0.4322... Generator Loss: 2.7270
Step: 70: Epoch 1/2... Discriminator Loss: 2.0954... Generator Loss: 0.2062
Step: 80: Epoch 1/2... Discriminator Loss: 1.2109... Generator Loss: 0.5417
Step: 90: Epoch 1/2... Discriminator Loss: 0.6478... Generator Loss: 1.8053
Step: 100: Epoch 1/2... Discriminator Loss: 0.7937... Generator Loss: 1.8659

Step: 110: Epoch 1/2... Discriminator Loss: 1.0040... Generator Loss: 0.7807
Step: 120: Epoch 1/2... Discriminator Loss: 1.0993... Generator Loss: 0.6176
Step: 130: Epoch 1/2... Discriminator Loss: 1.0465... Generator Loss: 0.7366
Step: 140: Epoch 1/2... Discriminator Loss: 1.5195... Generator Loss: 0.5006
Step: 150: Epoch 1/2... Discriminator Loss: 1.7786... Generator Loss: 0.2782
Step: 160: Epoch 1/2... Discriminator Loss: 1.3124... Generator Loss: 1.3335
Step: 170: Epoch 1/2... Discriminator Loss: 1.9887... Generator Loss: 0.1989
Step: 180: Epoch 1/2... Discriminator Loss: 1.1762... Generator Loss: 0.5372
Step: 190: Epoch 1/2... Discriminator Loss: 0.7833... Generator Loss: 1.3705
Step: 200: Epoch 1/2... Discriminator Loss: 1.1882... Generator Loss: 0.5366

Step: 210: Epoch 1/2... Discriminator Loss: 1.3384... Generator Loss: 0.6183
Step: 220: Epoch 1/2... Discriminator Loss: 0.8924... Generator Loss: 1.3869
Step: 230: Epoch 1/2... Discriminator Loss: 0.8386... Generator Loss: 0.9824
Step: 240: Epoch 1/2... Discriminator Loss: 0.9608... Generator Loss: 0.9323
Step: 250: Epoch 1/2... Discriminator Loss: 1.1202... Generator Loss: 0.6868
Step: 260: Epoch 1/2... Discriminator Loss: 0.8841... Generator Loss: 1.3923
Step: 270: Epoch 1/2... Discriminator Loss: 1.1626... Generator Loss: 0.5906
Step: 280: Epoch 1/2... Discriminator Loss: 1.1925... Generator Loss: 0.7229
Step: 290: Epoch 1/2... Discriminator Loss: 1.1850... Generator Loss: 0.4924
Step: 300: Epoch 1/2... Discriminator Loss: 1.1730... Generator Loss: 0.6080

Step: 310: Epoch 1/2... Discriminator Loss: 0.8728... Generator Loss: 0.8168
Step: 320: Epoch 1/2... Discriminator Loss: 0.8000... Generator Loss: 1.1347
Step: 330: Epoch 1/2... Discriminator Loss: 1.0292... Generator Loss: 1.8911
Step: 340: Epoch 1/2... Discriminator Loss: 0.8294... Generator Loss: 1.7329
Step: 350: Epoch 1/2... Discriminator Loss: 0.9321... Generator Loss: 0.7571
Step: 360: Epoch 1/2... Discriminator Loss: 0.8653... Generator Loss: 0.9849
Step: 370: Epoch 1/2... Discriminator Loss: 1.4115... Generator Loss: 0.3375
Step: 380: Epoch 1/2... Discriminator Loss: 0.8729... Generator Loss: 0.9766
Step: 390: Epoch 1/2... Discriminator Loss: 0.7763... Generator Loss: 1.5937
Step: 400: Epoch 1/2... Discriminator Loss: 0.9853... Generator Loss: 0.6671

Step: 410: Epoch 1/2... Discriminator Loss: 0.9784... Generator Loss: 1.0173
Step: 420: Epoch 1/2... Discriminator Loss: 0.7301... Generator Loss: 1.0711
Step: 430: Epoch 1/2... Discriminator Loss: 0.8782... Generator Loss: 1.4331
Step: 440: Epoch 1/2... Discriminator Loss: 1.4468... Generator Loss: 0.3907
Step: 450: Epoch 1/2... Discriminator Loss: 0.8245... Generator Loss: 1.9624
Step: 460: Epoch 1/2... Discriminator Loss: 0.8171... Generator Loss: 0.9680
Step: 470: Epoch 1/2... Discriminator Loss: 0.8812... Generator Loss: 0.8533
Step: 480: Epoch 1/2... Discriminator Loss: 0.9482... Generator Loss: 1.8117
Step: 490: Epoch 1/2... Discriminator Loss: 1.2506... Generator Loss: 0.5463
Step: 500: Epoch 1/2... Discriminator Loss: 0.7401... Generator Loss: 0.9874

Step: 510: Epoch 1/2... Discriminator Loss: 1.0339... Generator Loss: 0.7091
Step: 520: Epoch 1/2... Discriminator Loss: 0.9588... Generator Loss: 2.0683
Step: 530: Epoch 1/2... Discriminator Loss: 0.7921... Generator Loss: 1.4968
Step: 540: Epoch 1/2... Discriminator Loss: 0.9978... Generator Loss: 0.6379
Step: 550: Epoch 1/2... Discriminator Loss: 0.8205... Generator Loss: 0.9941
Step: 560: Epoch 1/2... Discriminator Loss: 1.1775... Generator Loss: 2.0490
Step: 570: Epoch 1/2... Discriminator Loss: 0.8506... Generator Loss: 1.4198
Step: 580: Epoch 1/2... Discriminator Loss: 1.0559... Generator Loss: 0.6699
Step: 590: Epoch 1/2... Discriminator Loss: 0.9265... Generator Loss: 2.8819
Step: 600: Epoch 1/2... Discriminator Loss: 1.2706... Generator Loss: 0.6076

Step: 610: Epoch 1/2... Discriminator Loss: 0.8690... Generator Loss: 0.9711
Step: 620: Epoch 1/2... Discriminator Loss: 1.3138... Generator Loss: 2.6223
Step: 630: Epoch 1/2... Discriminator Loss: 1.2197... Generator Loss: 0.5016
Step: 640: Epoch 1/2... Discriminator Loss: 1.0155... Generator Loss: 0.6796
Step: 650: Epoch 1/2... Discriminator Loss: 1.0755... Generator Loss: 0.6261
Step: 660: Epoch 1/2... Discriminator Loss: 0.9993... Generator Loss: 0.6869
Step: 670: Epoch 1/2... Discriminator Loss: 1.1707... Generator Loss: 2.1651
Step: 680: Epoch 1/2... Discriminator Loss: 0.6603... Generator Loss: 1.4774
Step: 690: Epoch 1/2... Discriminator Loss: 1.0311... Generator Loss: 0.7935
Step: 700: Epoch 1/2... Discriminator Loss: 0.6742... Generator Loss: 1.1329

Step: 710: Epoch 1/2... Discriminator Loss: 1.4525... Generator Loss: 0.3344
Step: 720: Epoch 1/2... Discriminator Loss: 1.2448... Generator Loss: 0.5364
Step: 730: Epoch 1/2... Discriminator Loss: 0.8164... Generator Loss: 1.1349
Step: 740: Epoch 1/2... Discriminator Loss: 0.9197... Generator Loss: 0.7774
Step: 750: Epoch 1/2... Discriminator Loss: 0.7362... Generator Loss: 1.5518
Step: 760: Epoch 1/2... Discriminator Loss: 0.4900... Generator Loss: 1.9430
Step: 770: Epoch 1/2... Discriminator Loss: 0.6996... Generator Loss: 1.7019
Step: 780: Epoch 1/2... Discriminator Loss: 1.1802... Generator Loss: 0.7414
Step: 790: Epoch 1/2... Discriminator Loss: 1.5416... Generator Loss: 0.3322
Step: 800: Epoch 1/2... Discriminator Loss: 1.9294... Generator Loss: 0.1960

Step: 810: Epoch 1/2... Discriminator Loss: 0.6313... Generator Loss: 1.2845
Step: 820: Epoch 1/2... Discriminator Loss: 1.3525... Generator Loss: 0.4246
Step: 830: Epoch 1/2... Discriminator Loss: 0.9313... Generator Loss: 1.4890
Step: 840: Epoch 1/2... Discriminator Loss: 0.9811... Generator Loss: 0.6220
Step: 850: Epoch 1/2... Discriminator Loss: 1.1436... Generator Loss: 2.3278
Step: 860: Epoch 1/2... Discriminator Loss: 1.1506... Generator Loss: 0.5735
Step: 870: Epoch 1/2... Discriminator Loss: 1.2378... Generator Loss: 0.5093
Step: 880: Epoch 1/2... Discriminator Loss: 1.2923... Generator Loss: 0.4129
Step: 890: Epoch 1/2... Discriminator Loss: 1.3057... Generator Loss: 0.4209
Step: 900: Epoch 1/2... Discriminator Loss: 0.7100... Generator Loss: 0.9649

Step: 910: Epoch 1/2... Discriminator Loss: 0.7189... Generator Loss: 1.0357
Step: 920: Epoch 1/2... Discriminator Loss: 0.8451... Generator Loss: 0.9910
Step: 930: Epoch 1/2... Discriminator Loss: 1.2932... Generator Loss: 0.4425
Step: 940: Epoch 1/2... Discriminator Loss: 1.5328... Generator Loss: 0.3157
Step: 950: Epoch 1/2... Discriminator Loss: 0.6556... Generator Loss: 1.4013
Step: 960: Epoch 1/2... Discriminator Loss: 0.7724... Generator Loss: 1.9210
Step: 970: Epoch 1/2... Discriminator Loss: 0.9855... Generator Loss: 0.6809
Step: 980: Epoch 1/2... Discriminator Loss: 0.9327... Generator Loss: 2.3936
Step: 990: Epoch 1/2... Discriminator Loss: 0.8972... Generator Loss: 0.7425
Step: 1000: Epoch 1/2... Discriminator Loss: 1.2346... Generator Loss: 0.4627

CelebA¶

Run your GANs on CelebA. It will take around 20 minutes on the average GPU to run one epoch. You can run the whole epoch or stop when it starts to generate realistic faces.

batch_size = 32
z_dim = 200
learning_rate = 0.0002
beta1 = 0.3


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 2

celeba_dataset = helper.Dataset('celeba', glob(os.path.join(data_dir, 'img_align_celeba/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, celeba_dataset.get_batches,
          celeba_dataset.shape, celeba_dataset.image_mode)

(202599, 28, 28, 3)
model input evaluated
(?, 28, 28, 3)
training...
Step: 10: Epoch 1/2... Discriminator Loss: 0.4997... Generator Loss: 1.9838
Step: 20: Epoch 1/2... Discriminator Loss: 0.3435... Generator Loss: 3.7795
Step: 30: Epoch 1/2... Discriminator Loss: 0.6435... Generator Loss: 8.6140
Step: 40: Epoch 1/2... Discriminator Loss: 0.1062... Generator Loss: 3.6392
Step: 50: Epoch 1/2... Discriminator Loss: 0.4228... Generator Loss: 5.2021
Step: 60: Epoch 1/2... Discriminator Loss: 0.4502... Generator Loss: 5.0158
Step: 70: Epoch 1/2... Discriminator Loss: 2.0957... Generator Loss: 0.2898
Step: 80: Epoch 1/2... Discriminator Loss: 1.2794... Generator Loss: 3.0261
Step: 90: Epoch 1/2... Discriminator Loss: 1.1259... Generator Loss: 0.5759
Step: 100: Epoch 1/2... Discriminator Loss: 1.3504... Generator Loss: 0.3874

Step: 110: Epoch 1/2... Discriminator Loss: 0.6998... Generator Loss: 1.1671
Step: 120: Epoch 1/2... Discriminator Loss: 0.8440... Generator Loss: 0.6740
Step: 130: Epoch 1/2... Discriminator Loss: 0.9744... Generator Loss: 5.0050
Step: 140: Epoch 1/2... Discriminator Loss: 0.5163... Generator Loss: 3.1381
Step: 150: Epoch 1/2... Discriminator Loss: 0.4798... Generator Loss: 3.8234
Step: 160: Epoch 1/2... Discriminator Loss: 0.5925... Generator Loss: 1.3277
Step: 170: Epoch 1/2... Discriminator Loss: 0.8627... Generator Loss: 0.7819
Step: 180: Epoch 1/2... Discriminator Loss: 1.2056... Generator Loss: 0.5139
Step: 190: Epoch 1/2... Discriminator Loss: 1.0239... Generator Loss: 0.7396
Step: 200: Epoch 1/2... Discriminator Loss: 0.8827... Generator Loss: 1.7161

Step: 210: Epoch 1/2... Discriminator Loss: 0.5807... Generator Loss: 1.3270
Step: 220: Epoch 1/2... Discriminator Loss: 0.8695... Generator Loss: 3.2031
Step: 230: Epoch 1/2... Discriminator Loss: 0.4783... Generator Loss: 2.4526
Step: 240: Epoch 1/2... Discriminator Loss: 0.8261... Generator Loss: 2.1026
Step: 250: Epoch 1/2... Discriminator Loss: 0.7498... Generator Loss: 0.9432
Step: 260: Epoch 1/2... Discriminator Loss: 3.0819... Generator Loss: 0.0659
Step: 270: Epoch 1/2... Discriminator Loss: 1.2674... Generator Loss: 0.7091
Step: 280: Epoch 1/2... Discriminator Loss: 1.0492... Generator Loss: 0.6669
Step: 290: Epoch 1/2... Discriminator Loss: 1.3501... Generator Loss: 0.4226
Step: 300: Epoch 1/2... Discriminator Loss: 0.5541... Generator Loss: 2.1671

Step: 310: Epoch 1/2... Discriminator Loss: 1.0192... Generator Loss: 2.6886
Step: 320: Epoch 1/2... Discriminator Loss: 0.8168... Generator Loss: 0.7952
Step: 330: Epoch 1/2... Discriminator Loss: 0.4962... Generator Loss: 1.3083
Step: 340: Epoch 1/2... Discriminator Loss: 0.6783... Generator Loss: 1.4004
Step: 350: Epoch 1/2... Discriminator Loss: 1.1734... Generator Loss: 0.5796
Step: 360: Epoch 1/2... Discriminator Loss: 1.0075... Generator Loss: 1.5461
Step: 370: Epoch 1/2... Discriminator Loss: 0.6809... Generator Loss: 1.0452
Step: 380: Epoch 1/2... Discriminator Loss: 1.5933... Generator Loss: 0.3298
Step: 390: Epoch 1/2... Discriminator Loss: 1.1347... Generator Loss: 0.7607
Step: 400: Epoch 1/2... Discriminator Loss: 0.8720... Generator Loss: 0.9170

Step: 410: Epoch 1/2... Discriminator Loss: 0.9963... Generator Loss: 0.7256
Step: 420: Epoch 1/2... Discriminator Loss: 0.3786... Generator Loss: 1.5072
Step: 430: Epoch 1/2... Discriminator Loss: 1.7778... Generator Loss: 0.2715
Step: 440: Epoch 1/2... Discriminator Loss: 1.2455... Generator Loss: 3.4792
Step: 450: Epoch 1/2... Discriminator Loss: 0.4129... Generator Loss: 2.7056
Step: 460: Epoch 1/2... Discriminator Loss: 0.5607... Generator Loss: 1.3412
Step: 470: Epoch 1/2... Discriminator Loss: 0.6043... Generator Loss: 1.7224
Step: 480: Epoch 1/2... Discriminator Loss: 0.7410... Generator Loss: 3.1676
Step: 490: Epoch 1/2... Discriminator Loss: 0.9996... Generator Loss: 0.7726
Step: 500: Epoch 1/2... Discriminator Loss: 0.9111... Generator Loss: 0.7123

Step: 510: Epoch 1/2... Discriminator Loss: 0.7561... Generator Loss: 0.9134
Step: 520: Epoch 1/2... Discriminator Loss: 0.6594... Generator Loss: 2.8341
Step: 530: Epoch 1/2... Discriminator Loss: 1.0317... Generator Loss: 2.9995
Step: 540: Epoch 1/2... Discriminator Loss: 0.9261... Generator Loss: 0.7586
Step: 550: Epoch 1/2... Discriminator Loss: 0.7592... Generator Loss: 1.2401
Step: 560: Epoch 1/2... Discriminator Loss: 0.5903... Generator Loss: 2.2362
Step: 570: Epoch 1/2... Discriminator Loss: 1.1759... Generator Loss: 0.6554
Step: 580: Epoch 1/2... Discriminator Loss: 0.5375... Generator Loss: 1.3791
Step: 590: Epoch 1/2... Discriminator Loss: 0.7316... Generator Loss: 1.0918
Step: 600: Epoch 1/2... Discriminator Loss: 1.3931... Generator Loss: 0.4960

Step: 610: Epoch 1/2... Discriminator Loss: 1.0899... Generator Loss: 0.5505
Step: 620: Epoch 1/2... Discriminator Loss: 0.5810... Generator Loss: 1.7206
Step: 630: Epoch 1/2... Discriminator Loss: 0.6751... Generator Loss: 1.4136
Step: 640: Epoch 1/2... Discriminator Loss: 2.3455... Generator Loss: 0.1259
Step: 650: Epoch 1/2... Discriminator Loss: 0.6086... Generator Loss: 2.3409
Step: 660: Epoch 1/2... Discriminator Loss: 0.8692... Generator Loss: 1.3613
Step: 670: Epoch 1/2... Discriminator Loss: 0.9916... Generator Loss: 0.6307
Step: 680: Epoch 1/2... Discriminator Loss: 0.9003... Generator Loss: 1.3840
Step: 690: Epoch 1/2... Discriminator Loss: 0.7162... Generator Loss: 1.0804
Step: 700: Epoch 1/2... Discriminator Loss: 0.6555... Generator Loss: 1.0291

Step: 710: Epoch 1/2... Discriminator Loss: 0.9784... Generator Loss: 2.6158
Step: 720: Epoch 1/2... Discriminator Loss: 0.8840... Generator Loss: 3.6204
Step: 730: Epoch 1/2... Discriminator Loss: 1.9459... Generator Loss: 0.2434
Step: 740: Epoch 1/2... Discriminator Loss: 1.5169... Generator Loss: 0.3489
Step: 750: Epoch 1/2... Discriminator Loss: 0.5309... Generator Loss: 1.5972
Step: 760: Epoch 1/2... Discriminator Loss: 1.1872... Generator Loss: 3.0359
Step: 770: Epoch 1/2... Discriminator Loss: 0.8313... Generator Loss: 1.9659
Step: 780: Epoch 1/2... Discriminator Loss: 1.7814... Generator Loss: 0.2700
Step: 790: Epoch 1/2... Discriminator Loss: 1.2668... Generator Loss: 0.4575
Step: 800: Epoch 1/2... Discriminator Loss: 1.2323... Generator Loss: 0.5113

Step: 810: Epoch 1/2... Discriminator Loss: 0.9094... Generator Loss: 2.9575
Step: 820: Epoch 1/2... Discriminator Loss: 0.6808... Generator Loss: 1.3976
Step: 830: Epoch 1/2... Discriminator Loss: 1.2741... Generator Loss: 0.4186
Step: 840: Epoch 1/2... Discriminator Loss: 0.6662... Generator Loss: 1.0747
Step: 850: Epoch 1/2... Discriminator Loss: 0.8566... Generator Loss: 0.8335
Step: 860: Epoch 1/2... Discriminator Loss: 0.5323... Generator Loss: 1.7929
Step: 870: Epoch 1/2... Discriminator Loss: 0.6378... Generator Loss: 0.9955
Step: 880: Epoch 1/2... Discriminator Loss: 0.8224... Generator Loss: 0.8040
Step: 890: Epoch 1/2... Discriminator Loss: 1.4847... Generator Loss: 0.3762
Step: 900: Epoch 1/2... Discriminator Loss: 1.4578... Generator Loss: 0.4009

Step: 910: Epoch 1/2... Discriminator Loss: 1.0317... Generator Loss: 0.6516
Step: 920: Epoch 1/2... Discriminator Loss: 0.5765... Generator Loss: 1.3809
Step: 930: Epoch 1/2... Discriminator Loss: 1.0265... Generator Loss: 2.6218
Step: 940: Epoch 1/2... Discriminator Loss: 1.1589... Generator Loss: 2.0189
Step: 950: Epoch 1/2... Discriminator Loss: 0.8908... Generator Loss: 0.7667
Step: 960: Epoch 1/2... Discriminator Loss: 0.5660... Generator Loss: 2.3845
Step: 970: Epoch 1/2... Discriminator Loss: 0.8342... Generator Loss: 0.8138
Step: 980: Epoch 1/2... Discriminator Loss: 1.1979... Generator Loss: 3.6667
Step: 990: Epoch 1/2... Discriminator Loss: 1.0894... Generator Loss: 0.5837
Step: 1000: Epoch 1/2... Discriminator Loss: 0.5123... Generator Loss: 1.7190

Submitting This Project¶

When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_face_generation.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.